Tamer Tahamoqa

Hi 55renchen, Could you clarify? Do you mean the Semi-Hard Negative triplet selection method? This blog post by Olivier Moindrot may be helpful: https://omoindrot.github.io/triplet-loss#triplet-mining

I would say it is Online since you are only selecting the triplets in a batch that pass the hard-negatives triplet selection condition instead of pre-computing the number of triplets...

Hi intergalacticSpaceDog, Yeah my bad. The '.to()' method is a Torch Tensor method but the img was not preprocessed to be turned into a Torch Tensor yet. I have fixed...

Hello 2017234312, I have managed to recreate this issue by using a Python 2.7 conda environment. I'm afraid you would have to use Python 3.6+ in order to use PyTorch...

Hi Daniel, Thank you very much for catching this one. The intention was only to speed up the triplet generation process and not to re-replicate the generated triplets across the...

Hello @eeric, Could you clarify more? Are you fully using CPU because a GPU resource is not available? The torch.utils.data.distributed.DistributedSampler() or the insightface dataset implementation may provide better peformance with...

Hello JinhaSong, Did you specify the input image to be 299x299? I think the minimum should be 150x150 but I am not sure on that to be honest. For reference,...

Hello @eeric, I set the normalization parameters as per the calculated training dataset RGB mean and standard deviation values in this [script](https://github.com/tamerthamoqa/facenet-pytorch-glint360k/blob/master/datasets/calculate_glint360k_rgb_mean_std.py) . It may or not be a better...

Hello htaufik, I am not sure to be honest, does ```cap = WebcamVideoStream(src="rtsp://192.168.150.79:11083").start()``` work?

Hello yonglianglan, Could you explain more on the steps you have taken to produce this issue? Since you first would have to upload an image to extract a face embedding...